MisleadingCharts
Back to the gallery

The 25 million boxes a log axis drew taller than the 92 million

Showing the misleading chart

Be the first to star this exhibit

A century of Florida’s orange crop, 101 seasons straight from USDA, on a log axis chosen for the honest reason that the crop has run from 8.65 million to 244 million boxes. Read off that axis, the fall into the season of Hurricanes Ian and Nicole is twice the height of the 2004 hurricanes’ fall, and a planning slide we drew concludes Ian took twice the bite. A log axis draws a fall’s percentage, not its size: Ian’s season lost 25.4 million boxes and 2004’s lost 92.2 million, 3.6 times as many.

01The claim

Read off a century of USDA orange counts on a log axis, the fall into Hurricane Ian’s season is twice the height of the fall after the 2004 hurricanes, so Ian took twice the bite out of Florida’s orange crop and supply cover should be sized to two 2004s.

02The trick

Every figure on the slide is USDA’s, and the log axis was a reasonable choice. The trouble is what a log axis does to a drop. It spaces equal ratios equally, so the height of a fall is set by its percentage and not by how much was lost. Ian’s season took the crop from 41.2 million boxes to 15.8 million, a fall of 61.6%. The 2004 season took it from 242.0 million to 149.8 million, a fall of 38.1%. In log terms the first ratio is 1.996 times the second, so the chart draws it twice as tall. In fruit it runs the other way: 25.4 million boxes against 92.2 million, so 2004 lost 3.6 times as much. The eye reads distance as amount, and on this axis distance is a ratio. So the slide’s ranking of supply shocks is a ranking by share, and as a ranking by share it is correct: 2022–23 is the worst one-season fall in the century. Ranked by boxes, which is what a juice blender has to replace, it is eleventh of 46 falls. The same axis pairs up falls that are nothing alike. 2024–25 lost 32.0% of the crop, 5.8 million boxes; 1962–63, the season of the December 1962 freeze, lost 33.4%, 36.3 million. The slide draws them at 0.80 and 0.85 of the 2004 height, near-twins six times apart in fruit. The recommendation follows the picture: a buffer sized to “two 2004s” is sized to about 184 million boxes, from a state whose whole 2025–26 crop was 12.95 million. (The slide is our own demonstration in the style of a juice blender’s planning deck. The company and its read-out are invented; the production figures are USDA’s.)

03The fix

The redraw puts the same 101 seasons on a linear axis from zero and outlines what each named season lost. 2004–05’s loss is a tall red box, 92.2 million boxes. Ian’s 25.4 million is a short one at the right, beside Irma’s 23.8 million. Two ranked panels then answer the two questions separately. By boxes, 2004–05 leads at 92.2 million, ahead of 1998–99’s 58.0 and 1981–82’s 46.6, and 2022–23 comes eleventh. By share of the crop, 2022–23 leads at 61.6% and 2004–05 is third at 38.1%. Both rankings are true; they answer different questions. For planning, the useful number is what an Ian-sized season would cost now: 61.6% of 2025–26’s 12.95 million boxes is 8.0 million, under a tenth of what the 2004 season took. Florida’s crop is now 5.3% of its 1997–98 peak, so no fall it can still have would match 2004 in boxes. The fix is to choose the axis by the question. A ratio question (how hard was each season hit?) belongs on a log axis or a chart of percent changes. An amount question (how much fruit do we need to replace?) belongs on a linear axis from zero, with the losses labelled in boxes. When a slide makes claims of both kinds, draw both charts and say which question each one answers. One caveat: hurricanes and freezes are not the only cause. Citrus greening, found in Florida in 2005, has been shrinking the crop since the mid-2000s, so not every box lost in the seasons since then was lost to the storm.